AI programming by children

The idea of children constructing artificial intelligence programs goes back to the early days of Logo (Papert Kahn 1975; Kahn 1977). After decades of little activity recent efforts to support students in making AI programs has come from Stephen Wolfram (Wolfram 2017b), Google (Google 2018a, 2018b), the Machine Learning for Kids website (Lane 2018), and the eCraft2Learn project (eCraft2Learn 2018b). Two technological developments underlie the feasibility of these efforts: (1) AI cloud services and (2) mainstream laptops and desktop computers that now can run sophisticated machine learning algorithms. And all of these developments can be made accessible in a web browser, thereby running on many platforms without the need to install software. Given appropriate programming tools children can make apps and intelligent robots that rely upon speech and image recognition services. They can add custom capabilities to their programs by using machine learning training and prediction. In doing so they may learn about perception, language, psychology, and the latest empowering technologies. We describe the addition of new programming blocks to the Snap! visual programming language (Harvey & Mönig 2010) that provide easy-to-use interfaces to both AI cloud services and deep learning functionality. Interactive learning materials have been developed and preliminarily trialled with students. We anticipate in future trials to observe children creatively using these new blocks to build very impressive programs. Children are likely to be even more motivated to program when the results are such capable programs.